WarpMap

Defines the VPIWarpMap object and related functions. More...

Data Structures
struct	VPIWarpGrid
	Holds VPI's warp grid definition. More...
struct	VPIWarpMap
	Defines the mapping between input and output images' pixels. More...

Functions
VPIStatus	vpiWarpMapAllocData (VPIWarpMap *warpMap)
	Allocates the warp map's control point array for a given warp grid. More...
void	vpiWarpMapFreeData (VPIWarpMap *warpMap)
	Deallocates the warp map control points allocated by vpiWarpMapAllocData. More...
VPIStatus	vpiWarpMapGenerateIdentity (VPIWarpMap *warpMap)
	Fills the given warp map with an identity mapping. More...
#define	VPI_WARPGRID_MAX_HORIZ_REGIONS_COUNT   4
	Maximum number of regions horizontally in a warp grid.
#define	VPI_WARPGRID_MAX_VERT_REGIONS_COUNT   4
	Maximum number of regions vertically in a warp grid.
#define	VPI_WARPGRID_MIN_REGION_WIDTH   64
	Minimum warp grid region width.
#define	VPI_WARPGRID_MIN_REGION_HEIGHT   16
	Minimum warp grid region height.

Detailed Description

Defines the VPIWarpMap object and related functions.

Warp map holds the mapping from output to input images, used in Remap algorithm. It allows both dense and sparse mapping.

---

Data Structure Documentation

VPIWarpGrid

struct VPIWarpGrid

Holds VPI's warp grid definition.

This structure defines the layout of the control points in the destination image of a remap operation.

The control points are used as the basis for geometric transformation from source image to destination image. The remaining points are transformed based on the interpolation. Thus the density of the control points controls the quality of the geometric transformation.

This is an example of defining regions in the image:

           warp grid width
     /                         \
    /                           \
   /                             \
 
regionWidth[0]    regionWidth[numHorizRegions-1]
   /    \                  /    \
  |------|                |------|
  --------------------------------                                 \
  |******* *******        *******|--                                \
  |* . . * *     *        *     *|  \                                \
  |* . . * *     *  ...   *     *|   regionHeight[0]                  \
  |* . . * *     *        *     *|  /                                  \
  |******* *******        *******|--                                    \
  |                              |                                       \
  |   .       .              .   |                                  
  |   .       .              .   |                                         warp grid height
  |   .       .              .   |
  |                              |                                       /
  |******* *******        *******|--                                    /
  |*     * *     *        *     *|  \                                  /
  |*     * *     *  ...   *     *|   regionHeight[numVertRegions-1]   /
  |*     * *     *        *     *|  /                                /
  |******* *******        *******|--                                /  
  --------------------------------                                 /

This is an example of defining control points in one region:

 *********
 *  +  + *-- \
 *       *    vertInterval
 *  +  + *-- /
 *       *
 *********
    |--|
horizInterval

Here's an example of a WxH dense grid definition:

VPIWarpGrid grid;
grid.numHorizRegions  = 1;
grid.numVertRegions   = 1;
grid.horizInterval[0] = 1;
grid.vertInterval[0]  = 1;
grid.regionWidth[0]   = W;
grid.regionHeight[0]  = H;

Restrictions

• numHorizRegions cannot exceed VPI_WARPGRID_MAX_HORIZ_REGIONS_COUNT.
• numVertRegions cannot exceed VPI_WARPGRID_MAX_VERT_REGIONS_COUNT.
• Intervals must be power-of-two.
• Alignment restrictions:
  • regionWidth[0] to regionWidth[numHorizRegions-2] must be aligned to VPI_WARPGRID_MIN_REGION_WIDTH and at least VPI_WARPGRID_MIN_REGION_WIDTH.
  • regionWidth[numHorizRegions-1] must be at least VPI_WARPGRID_MIN_REGION_WIDTH.
  • regionHeight[0] to regionHeight[numVertRegions-2] must be aligned to VPI_WARPGRID_MIN_REGION_HEIGHT and at least VPI_WARPGRID_MIN_REGION_HEIGHT.
  • regionHeight[numVertRegions-1] must be at least VPI_WARPGRID_MIN_REGION_HEIGHT.

Definition at line 156 of file WarpGrid.h.

Collaboration diagram for VPIWarpGrid:

Data Fields
int16_t	horizInterval[VPI_WARPGRID_MAX_HORIZ_REGIONS_COUNT]	Horizontal spacing between control points within a given region. Must be power-of-two.
int8_t	numHorizRegions	Number of regions horizontally.
int8_t	numVertRegions	Number of regions vertically.
int16_t	regionHeight[VPI_WARPGRID_MAX_VERT_REGIONS_COUNT]	Height of each region.
int16_t	regionWidth[VPI_WARPGRID_MAX_HORIZ_REGIONS_COUNT]	Width of each region.
int16_t	vertInterval[VPI_WARPGRID_MAX_VERT_REGIONS_COUNT]	Vertical spacing between control points within a given region. Must be power-of-two.

VPIWarpMap

struct VPIWarpMap

Defines the mapping between input and output images' pixels.

This structure is used as input to Remap. It defines the control point positions in the input image. The corresponding positions in the output image is implicitly defined by the warp grid definition.

Definition at line 87 of file WarpMap.h.

Collaboration diagram for VPIWarpMap:

Data Fields
VPIWarpGrid	grid	Warp grid control point structure definition. It implicitly defines the control point positions in the output image.
VPIKeypoint *	keypoints	Pointer to an array with control point positions in the input image corresponding to those in the output image. Coordinates are absolute, (0,0) is the top/left corner of output image.
int16_t	numHorizPoints	Number of points horizontally. Must match the number of points defined by the grid. This is calculated by vpiWarpMapAllocData.
int16_t	numVertPoints	Number of points vertically. Must match the number of points defined by the grid. This is calculated by vpiWarpMapAllocData.
int32_t	pitchBytes	Number of bytes between one control point and the one immediately below. Must be at least sizeof(VPIKeypoint)*numHorizPoints.

Function Documentation

vpiWarpMapAllocData()

VPIStatus vpiWarpMapAllocData

(

VPIWarpMap *

warpMap

)

#include <vpi/WarpMap.h>

Allocates the warp map's control point array for a given warp grid.

This function will read the warp grid structure and allocated an appropriately sized control point array, filling the numHorizPoints, numVertPoints, strideBytes and keypoints attributes of VPIWarpMap. The warp map must be deallocated by vpiWarpMapFreeData when no longer needed.

Parameters

[in,out]

warpMap

The warp map whose keypoint array will be allocated. The warp grid attribute must be already filled out with the grid structure. keypoints must be NULL. All attributes other then grid and keypoints will be overwriten.

Returns

VPI_SUCCESS if allocation succeeded, or some other VPIStatus value in case of error.

vpiWarpMapFreeData()

void vpiWarpMapFreeData

(

VPIWarpMap *

warpMap

)

#include <vpi/WarpMap.h>

Deallocates the warp map control points allocated by vpiWarpMapAllocData.

This function does nothing if control points array is NULL. It sets numHorizPoints, numVertPoints and strideBytes to zero, and keypoints to NULL.

Parameters

[in,out]

warpMap

Warp map whose control points array needs to be deallocated.

vpiWarpMapGenerateIdentity()

VPIStatus vpiWarpMapGenerateIdentity

(

VPIWarpMap *

warpMap

)

#include <vpi/WarpMap.h>

Fills the given warp map with an identity mapping.

This function is useful if the user wants to specify their own mapping. It sets the control points coordinates to the destination coordinates as defined implicitly by the warp grid. The user then can iterate through these points and apply a custom mapping function to each one.

Parameters

[in,out]

warpMap

Warp map with allocated control point array to be filled with identity mapping.

Returns

VPI_SUCCESS if allocation succeeded, or some other VPIStatus value in case of error.